Compressible Fluids Notes

150
15.Method of Characteristics: Two-
Dimensional Steady Flow
So far, we have only dealt with flows that are one-dimensional (Chapters 1-11)
or are comprised of simple two-dimensional elements such as oblique shocks
or Prandtl-Meyer expansion fans (Chapters 12-14). For more complex flows
(say, over a curved wing or fuselage), it becomes necessary to develop new
methodologies that can solve general two dimensional or multi-dimensional
flows. One of the most versatile approaches is called the method of
characteristics. It uses the fact that compressible flow is controlled by
disturbances that propagate along waves or lines (also called characteristics or
characteristic lines). These are not new concepts; we have already encountered
the envelope of disturbances called the Mach angle in Section4.3. As will be
discussed in this chapter, the Mach line is a characteristic in steady, two-
dimensional supersonic flow. By tracking how lines of influence propagate at
the local Mach angle through a flow, it is possible to solve the entire flow field;
this is the essence of the method of characteristics.
The method of characteristics is very general and very powerful. It can either
be implemented on an ad hoc basis by hand or systemized into a computer-
based algorithm. Implementing it for complex flows with shock waves,
however, it can become cumbersome, and it has been largely replaced by
finite-difference computational fluid dynamics (CFD) in recent decades. The
method of characteristics is still worthy of our attention: examining
characteristics is often essential to obtain a feel for how a flow responds to
boundary conditions. In addition, the development of modern CFD codes is
often built on an understanding of how characteristics control a compressible
flow. Interestingly, some modern developments in CFD (such as space-time
finite elements) are remarkably similar to the classical method of
characteristics we will develop here.
15.1 Governing Equations for Irrotational Flow
In this Chapter, we will limit our attention to two-dimensional, steady
supersonic flow. In addition, we will only consider irrotational flow. This is
an important assumption, and it is worthwhile to detail what types of flow can,
and cannot, be considered irrotational. An irrotational flow is a flow in which
the vorticity of the flow is everywhere zero. Vorticity is defined as the curl of
the velocity vector: vorticity = , where is the velocity vector and is
the grad operator ( for two-dimensional Cartesian coordinates).
Swirling flows with eddies are obviously flows with vorticity. Some less
obvious examples of rotational flow are flows in boundary layers or flows
behind curved shock waves
*
; the version of the method of characteristics
developed here will not be applicable to these flows. An example of an
irrotational flow (where =0 everywhere in the flow) includes an initially
uniform flow passing through an isentropic converging-diverging nozzle.
The fact that the curl of velocity is everywhere zero in an irrotational flow
means that the velocity vector can be expressed in terms of a potential
function :
Recall that a potential function is a scalar function whose gradient gives a
vector field. It is a compact way of expressing the velocity vector field, and
since it replaces the components of velocity with a single function, can lead to
simplification in solving the flow. For example, the conservation of mass
(continuity) for a steady flow:
*
It may not be obvious why flows with curved shock waves have vorticity. However, a
powerful principle knows as Croccos theorem, expressed as ,
explicitly links vorticity with a gradient in entropy. Since a curved shock wave in an
initially uniform flow produces a gradient in entropy as the strength of the shock
changes, the flow downstream of the shock must have non-zero vorticity.
Implementing the method of characteristics for rotational flows (such as behind curved
shocks) necessitates tracking an additional characteristic to account for the changing in
entropy through the flow.
V V
i

+
V V ( ) Ts =
V
V
V =
V ( ) 0 =
Compressible Fluids Notes
151
(note that this is simply the vector form of the conservation of mass derived in
Section 2.1) can be written using the potential function as
(15.1)
We can eliminate (density) completely by invoking the momentum equation
(recall Section3.2):
or
(15.2)
If we further limit ourselves to isentropic flow, we can relate changes in
density to changes in pressure by the definition of the speed of sound (recall
Section4.1):
Thus, at any point in an isentropic flow:
Using (15.2) in this result, we obtain:
Thus, we can express the gradient of density in the x-direction as
and likewise for the y-direction:
We can use these expressions in (15.1) to eliminate density entirely:
This relation is called the velocity potential equation and represents both the
continuity and momentum equations for irrotational, isentropic flow. This
equation is still coupled to the energy equation via the sound speed c. Since
sound speed is a function of temperature, it will be necessary to solve for
temperature using the energy equation.
15.2 Properties of Hyperbolic Systems:
Characteristics
We will re-introduce the velocity components u and v for a moment to make a
few points:
(15.3)
Thus, compressible flow is governed by a PDE of the form:
x

u ( )
y

v ( ) +
x

x
( )
y

y
( ) + 0 = =

xx

yy
+ ( )
x
x

y
y

+ + 0 =
dp

------ VdV + 0 =
dp
d V
2
( )
2
--------------
d
x
2

y
2
+ ( )
2
------------------------- = =
c
2

p


s
=
d
dp
c
2
------ =
d

c
2
-----d

x
2
2
-----

y
2
2
----- +



=
x

c
2
-----
x

x
2

y
2
+
2
-----------------

c
2
-----
x

xx

y

yx
+ ( ) = =
y

c
2
-----
y

x
2

y
2
+
2
-----------------

c
2
-----
x

xy

y

yy
+ ( ) = =
1

x
2
c
2
-------

xx
2
x

y
c
2
--------------
xy
1

y
2
c
2
-------

yy
+ 0 =
1
u
2
c
2
-----

xx
2uv
c
2
---------
xy
1
v
2
c
2
-----

yy
+ 0 =
Compressible Fluids Notes
152
where , , and . From the basic theory of partial
differential equations, PDEs of this form can be classified as follows:
The canonical elliptic PDE is the Laplace equation: it is governed by smooth
solutions where all boundary conditions have an influence on the solution at
every point in the domain (think of the steady state temperature distribution in
a plate with temperature prescribed along the edges). The classic example of a
parabolic equation is the heat (or diffusion) equation: in parabolic systems,
initial conditions simply spread or diffuse outward over time, but the influence
of a source is felt instantaneously throughout the domain (think of temperature
diffusing outward into a rod after being given an initial heat pulse at one end).
The most familiar example of a hyperbolic equation is the wave equation
( ), where information propagates via waves that move at a
specific speed. With a hyperbolic system, a source or disturbance has a clearly
defined zone of influence (return to the picture of throwing rocks into a pond
or river, as discussed in Section 4.3).
For steady, irrotational flow, the quantity can be shown to equal
Thus, the type of PDE governing a flow changes as the flow goes from
subsonic to supersonic. Subsonic flow is governed by an elliptic PDE: since
the flow is everywhere subsonic, the influence of upstream and downstream
conditions can be felt everywhere in the flow, and the solution is able to
accommodate smoothly. In fact, in the limit as V <<c (incompressible limit),
the potential equation reverts to the Laplace equation ( ). In
supersonic flow, the effect of a boundary or source is only felt in a domain of
influence that is defined by the speed at which information propagates. The
fact that every point in the solution cannot communicate to everywhere else in
the domain can give rise to discontinuous jumps (shock waves).
Hyperbolic partial differential equations in general have a unique feature.
Along certain lines, called characteristics, the partial differentials become
undefined (or, more properly, indeterminate). Along these lines the PDE may
collapse to simpler ordinary differential equations or even algebraic
equations. You are encouraged to consult a text book on partial differential
equations for a full development; the development here will only give you a
taste of the idea. For example, suppose we solve (15.3) for :
(15.4)
Thus, is known as a function of how u and v vary along the x-direction (
and ). This partial derivative will be indeterminate, however, when v =c
(denominator goes to zero). Consider a point in the flow where this happens.
The angle the flow makes with the horizontal is:
A
x
2
2

2B
xy
2

C
y
2
2

+ + 0 =
A 1
u
2
c
2
----- = B
uv
c
2
----- - = C 1
v
2
c
2
----- =
B
2
AC 0 elliptic <
B
2
AC 0 parabolic =
B
2
AC 0 hyperbolic >
t
2
2

c
2
x
2
2

0 =
B
2
AC
B
2
AC
u
2
v
2
+
c
2
----------------- 1
V
2
c
2
------ 1 M
2
1 = = =
x
2
2

y
2
2

+ 0 =
y
v
y
v

yy
2uv
c
2
---------
x
v
1
u
2
c
2
-----



x
u

1
v
2
c
2
-----



------------------------------------------------ = =
y
v
x
u
x
v
Compressible Fluids Notes
153
Thus, when the angle of the flow with respect to
the horizontal becomes equal to the Mach angle,
the partial derivative of the v-velocity component in the y-direction becomes
undefined. The only way the expression (15.4) can avoid blowing up in this
case is if the numerator is also zero, and the expression is indeterminate.
Setting the denominator of (15.4) equal to zero, we obtain the compatibility
equation:
along the x-direction.
This equation is really an ordinary differential equation, and we should replace
with , since it is only dependent on how properties vary in a single
direction (the x-direction). The horizontal line in this case becomes a
characteristic.
Note that the orientation of the x and y axes is arbitrary; we can always rotate
the axes to find an orientation where one axis will be at the Mach angle with
respect to the flow. Thus, any line that is at the Mach angle with respect to
the flow is a characteristic.
The formal technique to find the characteristics of a system is to express the
system as algebraic linear equations:
Solving for using the techniques of linear algebra (Cramers rule):
(15.5)
And likewise for and . Setting the denominator equal to zero gives a
quadratic equation for :
Thus, there are two different lines along which the partial derivative becomes
indeterminate, this expression gives the slope of these lines. Introducing
and and :
Note that , so this expression simplifies to:
V

v

=

c
u

c
V
--- sin
1
M
---- - sin = = =
2uv
c
2
---------
x
v
1
u
2
c
2
-----



x
u
0 =
x

x d
d
A
x
2
2

2B
xy
2

C
y
2
2

+ + 0 =
dx
x
u


dy
y
u


0
y
v


+ + du =
0
x
u


dx
y
u


dy
y
v


+ + dv =
x
u
x
u
0 2B C
du dy 0
dv dx dy
A 2B C
dx dy 0
0 dx dy
-------------------------------
d u 2Bdy Cdx ( ) dvCdy
Adydy dx 2Bdy Cdx ( )
------------------------------------------------------------------- = =
y
u
y
v
x d
dy
x d
dy


char
B B
2
AC
A
----------------------------------
uv
c
2
------
u
2
v
2
+
c
2
----------------- 1
1
u
2
c
2
-----
----------------------------------------------- = =
u V cos = v V sin = sin
1 1
M
---- - sin
1 c
V
--- = =
x d
dy


char
sin cos
sin
2

-----------------------
cos
2
sin
2
+
sin
2

--------------------------------- 1
1
cos
2

sin
2

--------------
--------------------------------------------------------------------------------- =
cos
2
sin
2
+
sin
2

--------------------------------- 1
1
tan
------------ =
x d
dy


char
+

( ) tan =
Compressible Fluids Notes
154
The graphical interpretation of this result is shown here:
Note than the expression gives the slope of a line that is
oriented at the Mach angle (or ) with respect to the flow direction. We
have encountered these lines before: they are Mach lines (see Section4.3).
We now see that Mach lines are also characteristics. We will denote the two
Mach lines as the right-running characteristic and the left-running
characteristic (C
I
and C
II
, respectively). These names derive from the fact that
if we disturb the flow at point P (imagine, dipping a finger into the flow),
would see one Mach line running off toward the right and another toward the
left. In the sketch here, the characteristics are shown to curve, because they
may be moving into a region of flow where and are changing. But locally
(at the point P), their slopes are given by .
We now turn to the numerator of (15.5), its value must also be zero in order to
keep the flow field derivatives from blowing up:
Using the fact that this condition will only apply along the characteristics,
where we know , we can obtain (note: many steps of algebra
skipped over):
Using and and considerable manipulation, we obtain:
This is our compatibility equation, where the sign applies along C
I
characteristics and the + sign applies along C
II
characteristics. This is an
ordinary differential equation, which we can now integrate. Looking back to
Section13.1, however, we note that we already integrated this relation: when
integrated, it yields the Prandtl-Meyer Function: (M). Thus, our
compatibility conditions become algebraic equations:
Thus, the quantities C
I
and C
II
are constant along right-running and left-
running characteristic, respectively. This concept takes some getting used to,
but is very powerful: imagine a point in a supersonic flow that generates a tiny
disturbance. The disturbance will propagate out from that point along Mach
lines (one to the right, one to the left), and they carry a piece of information
along with them. That information is that the quantity C
I
=(M) + is a
constant (or C
II
=(M) is a constant). As this disturbance propagates
through the flow, it will encounter regions with different flow angle and
different Mach number M. The sum of (M) +, however, will always remain
a constant (provided the flow remains supersonic and irrotational). This
remarkable property permits us to solve an entire supersonic flowfield,
provided we know the conditions entering the flow boundary.
V

+

x direction
y direction
C
I
C
II
point P
x d
dy


char
+

( ) tan =
+

( ) tan
dv
du
------
dx
dy
------
2B
C
------- =
dy
dx
------
dy
dx
----- -


char
=
dv
du
------
uv
c
2
------
u
2
v
2
+
c
2
----------------- 1 +

1
v
2
c
2
-----
------------------------------------------- =
u V cos = v V sin =
d M
2
1
dV
V
-------
+

=
C
I
=(M) + =constant along right-running characteristics
C
II
=(M) =constant along left-running characteristics
Compressible Fluids Notes
155
15.3 Solution Techniques
15.3.1Solving for a Point in the Flow
Let us assume that there are two points in a supersonic flow at which we have
complete knowledge of the flow: Points 1 and 2. At Point 1, the flow is
moving upward and to the right with flow angle =20 at Mach 2. At Point 2,
the flow is less inclined ( =5) but moving a bit faster (M =2.1). We will also
assume we know the exact (x,y) coordinates of each point.
We can imagine that there is an infinitesimal disturbance emanating from Point
1, that sends characteristic Mach lines streaming off to the left and right. Since
we know and Mach number at Point 1, we know the value of C
I
and C
II
these
characteristics carry with them: (C
I
)
1
=26.38 +20 =46.38 and
(C
II
)
1
=26.38 20 =6.38. Likewise, Point 2 has two characteristics
streaming away from it as well: (C
I
)
2
and (C
II
)
2
.
At some point in the flow, the right running characteristic from Point 1 (C
I
)
1
and the left running characteristic from Point 2 (C
II
)
2
will intersect. Note that
we can approximately locate this point in the flow by making a geometric
construction (i.e., straight edge and protractor): we know the right running and
left running characteristics have angles + and with respect to the
horizontal. While the characteristic may curve as they move through regions
of nonuniform flow, we can locally approximate them as straight lines.
Note that while we do not know a priori what the flow properties are at
Point 3, we do know that (C
I
)
1
=46.38 and (C
II
)
2
=24.07. Both these values
must apply at Point 3. Using this information, we can solve for
3
and
3
:
Knowing
3
, we can find Mach number using a table of Prandtl-Meyer
function values: M
3
=2.34. Thus, we can completely solve for the conditions
at Point 3 (pressure, temperature, etc. can be found using isentropic relations,
assuming we know the stagnation temperature and pressure of the flow).
Now that we know the values of and at Point 3 (
3
=11.15,
3
=25.30),
we could revise the geometric construction we used to find the (x,y) location of
Point 3 (i.e., get out protractor and straight edge again) by using the average
values of the flow and Mach angles:
, etc. for
23
,
13
,
23
since these values are more representative of the actual flow the characteristics
propagate through. Thus, we can improve the accuracy of our solution.
*
Note
that the values of the flow properties at Point 3 do not change as we do this
iteration, only the location of Point 3 changes.

C
I
C
II

2
------------------- =
C
I
C
II
+
2
------------------- =

3
46.38 24.07
2
-------------------------------------- 11.15 = =
3
46.38 24.07 +
2
-------------------------------------- 35.22 = =
V
1

1
=20

1
+
1
=50

1

1
=10
y direction
(C
I
)
2 =

2 +

2 =
3
3
.4
4

(
C
I
I
) 1
=

1
=
6
.
3
8

Point 1
M
1
=2.0

1
=20

1
=30

1
=26.38
V
2

2
=5

2
+
2
=33.44

2

2
=23.44
(C
I
)
1 =

1 +

1 =
46.38
(
C
II
) 2
=
2

2
=
2
4
.
0
7

Point 2
M
2
=2.1

2
=5

2
=28.44

2
=29.07
Point 3
x direction

13
1
2
---
1

3
+ ( ) 15.58 = =
Compressible Fluids Notes
156
Note that the Mach number at Point 3 is greater than at Points 1 and 2. This
agrees with the intuition we developed in studying one-dimensional flow:
since the flow is supersonic and diverging, it should accelerate and Mach
number should increase.
Also note that the characteristics do not terminate at Point 3, they continue on
through the flow, carrying their information about C
I
and C
II
with them.
Alternative, we can imagine that new C
I
and C
II
characteristics emanate from
Point 3; their values of (C
I
)
3
and (C
II
)
3
will be the same as (C
I
)
1
and (C
II
)
2
.
When they intersect characteristics of another type (e.g., a right running
characteristic intersects a left running characteristic), we can solve for that new
point in the flow via the same technique.
15.3.2Solving for a Point on the Wall
Let us return to the C
I
characteristic that emanated from Point 1. Let us
assume that this characteristic eventually encounters a wall. Again, we can
estimate the point of this intersection with the wall via a geometric
construction, assuming the characteristic remains at a constant angle
1
+
1
.
Where it hits the wall at Point 4, value of (C
II
)
1
must still apply. We do not
have the C
I
characteristic at the point, but since the flow must be parallel to the
wall, we do have a value of (say,
wall
=
4
=13). Thus, we can solve for
4
and Mach number:
So, M
4
=1.75, from a look-up of Prandtl-Meyer function values. Thus, we
now have complete knowledge about this point on the wall, including the value
of (C
I
)
4
for a new right-running characteristic that is emanated from this point.
15.4 Solving a Complete Flowfield
By combining the point-wise solution techniques developed above, we can
now march through a supersonic flow and solve for each point where
characteristics intersect. This requires that we have complete knowledge of the
flow along some line (not a characteristic line). Typically, this is an upstream
boundary where the flow enters the domain of interest. For a nozzle, for
example, we would need to know the flow conditions at some line or plane
downstream of the throat. Note that we cannot initialize our characteristic
solution with the sonic flow at the throat, since the method of characteristics
breaks down for sonic flow. We need to initialize the solution with a supersonic
flow (say, Mach 1.1).
Once we have an upstream boundary specified, we can start streaming
characteristics downstream from points on the boundary. Where the
characteristics intersect, we can solve for the flow conditions at that point. If
the characteristics intersect the wall, we will know the wall angle and thus can
*
In practice, it is often not worth correcting the position of the new point using the
average values. In a numerical solution, it is often easier to just add more
characteristics, thus shortening the distance over which they need to travel before
solving for the next point and thereby decreasing the error. By continuing to increase
the number of characteristics, you can quickly converge to a final solution.

4
C
II
( )
1

4
+ 19.38 = =
V
1

1
=20

1
+
1
=50

1

1
=10
y direction
(
C
I
I
) 1
=

1
=
6
.
3
8

Point 1
M
1
=2.0

1
=20

1
=30

1
=26.38
(C
I
)
1 =

1 +

1 =
46.38
x direction
Point 4

wall
=
4
=13
(C
I
)
4 =

4 +

4
Compressible Fluids Notes
157
solve for the flow there, and stream new characteristics back into the solution
domain.
Constructing a table of flow variables and characteristics is often essential to
successfully implement the method. This is illustrated in the numerical
example below.
15.5.1Numerical Example: Supersonic Flow into a
Converging Channel
Consider supersonic flow entering a channel that has converging walls of fixed
angle 10. The flow enters at Mach 3, and is parallel to the wall at the wall and
runs horizontal at the center line of the channel. Using the method of
characteristics, solve for the flow as it continues down the channel.
We will initialize our solution by choosing three points on the boundary that
was specified (the inlet plane of the channel): Points 1, 2, and 3. Since we
have complete knowledge about these points, we can determine the values of
C
I
and C
II
and stream those values into the flow. For example, the C
I
characteristic from Point 1 intersects the C
II
characteristic from Point 2 at Point
4. Knowing C
I
and C
II
, we solve for Point 4, and then stream characteristics
on from Point 4 to Points 6 and 7.
In the table shown here, values that are known a priori either from the initial
boundary or the wall are denoted with a heavy box. Values that are obtained by
streaming characteristics from known values are denoted with a double-lined
box. Once any two of the four parameters (C
I
, C
II
, , ) is known, the other
two are automatically determined, and then the Mach number and other flow
parameters (Mach angle, etc.) can be solved for. These other values are listed
in the table (M, , +, and ), but note that they are not required to solve
for the values of C
I
, C
II
, , or . They are used in performing the graphical
construction to determine where the points will lie in physical space. We have
stopped at Point 7, but the method can be extended further.
Note that solving for the lower half of the channel was redundant. Since the
channel was symmetric, we could have replaced the center line with a
horizontal wall and the solution would have remained the same. If we wish to
increase the accuracy of our solution, we can increase the number of initial
characteristics emitted from the initial datum line. For example, we can
introduce a point halfway between Points 1 and 2.
15.6 Breakdown of the Method
The solution above can be continued until one of two things happen:
Characteristics running in the same direction cross.
The flow becomes subsonic.
Point C
I
C
II
M +
1 39.757 59.757 -10 49.757 3.00 19.47 9.47 -29.47
2 49.757 49.757 0 49.757 3.00 19.47 19.47 -19.47
3 59.757 39.757 10 49.757 3.00 19.47 29.47 -9.49
4 39.757 49.757 -5 44.757 2.753 21.3 16.3 -26.3
5 49.757 39.757 5 44.757 2.753 21.3 26.3 -16.3
6 27.755 49.757 -10 39.757 2.527 23.31 13.31 -33.31
7 39.757 39.757
1
2
3
4
5
6
7
8
M
a
c
h

3
d
a
t
u
m

o
f

k
n
o
w
n

c
o
n
d
i
t
i
o
n
s
Compressible Fluids Notes
158
Characteristics running in the same
direction means, for example, two C
I
characteristics, streaming from different
points in space, intersect, as shown here.
When this happens, technically the two
waves merge to form a stronger, finite
amplitude wave that can no longer be
considered a characteristic; it has become
a shock wave (albeit a weak one). If you
review Chapter 13, this is exactly how
we discussed oblique shock waves forming on a smoothly curved corner:
Mach lines (or characteristics) intersect and coalesce to form an oblique shock.
Once a shock wave appears, our flow is no longer isentropic or irrotational, the
key assumptions upon which our method was developed. If the shock wave
generated is weak, it may be possible to get away with continuing to treat it
as a characteristic, although the fact that two characteristics have now become
one can wreak havoc on our bookkeeping system (or computerized algorithm).
If characteristics continue to merge, the oblique shock gets stronger, and we
really cannot justify continuing the solution. There are ways the method of
characteristics can be modified to handle these situations, but finite difference
CFD codes are typically much better in these cases.
The other thing that will cause our solution method to break down is the
appearance of sonic (or subsonic) flow. Again, there are ways that the method
of characteristics can be modified to handle interacting with regions of
subsonic flow (although the method itself cannot apply in those regions).
15.7 Some Comments
As one is implementing the method of characteristics, the operation using the
information propagated along characteristics to solve for the flow at a new
point where characteristics interact becomes routine, consisting of simple
addition and subtraction and table look-ups of (M) and (M). The main
challenge becomes careful bookkeeping of the characteristic information. As
simple as these operations are, one should not lose sight of the fact that, in
implementing the method of characteristics, you are actually solving a coupled
set of nonlinear partial differential equations! This is no small feat, since in
general, solving nonlinear PDEs is a task that demands considerable
mathematical prowess or serious computational power.
Another thought to keep in mind in implementing the method of characteristics
is that, to some degree, this is how Nature itself solves a compressible flow.
In actuality, there are an infinite number of characteristics in a real, physical
domain of flow: they form a continuum. In the method of characteristics,
these are approximated by a finite number of characteristics. In laboratory
experiments, the trajectory of a few characteristics can be observed by, for
example, making scratches on the walls of a channel to induce a finite (but still
very weak) disturbance that can be visualized by techniques such as Schlieren.
This has been done in the photograph shown here, where grooves that have
been scratched into the walls of a wind tunnel permit the characteristics to be
visualized. Characteristics are real. (The zebra-patterns here are not related to
characteristics; they are a result of a flow visualization technique called
interferometry that permits us to visualize density contours in a flow.)
Finally, if the development of the method of characteristics seems so
mathematical as to be opaque, then you might look forward to Chapter 16,
where the method is developed for one-dimensional unsteady flow. In one-
(
C
I )
1
(C
I
)
2
w
e
a
k

s
h
o
c
k
Taken from M. van Dyke, An Album of Fluid Motion, Parabolic Press,
Stanford, California, 1982. Originally from Hiller & Meier (1975).
Compressible Fluids Notes
159
dimensional unsteady flow, the characteristics again turn out to be acoustic
waves, but the characteristics and compatibility conditions can be derived in a
more transparent and intuitive fashion.
15.8 Examples of Applying the Method of
Characteristics
*
The method of characteristics can be implemented in a computer program such
that a large number of characteristics can be tracked. This has been done
using Matlab, and the results for an initially slightly supersonic flow
(M
in
=1.000001) entering a channel with an upper wall that diverges at a
constant angle of 20 are shown below. The method of characteristics solution
was initialized with a Prandtl Meyer expansion fan at the corner: knowing the
Prandtl-Meyer flow, a finite number of characteristics were specified. The
characteristics then reflected off the lower wall and proceeded to bounce
back and forth between the two walls, with each interaction between
characteristics being solved via the methods outlined above.
It is interesting to see how this two-dimensional solution compares with the
one-dimensional solution to isentropic flow in a diverging channel developed
in Chapter 5. This comparison is made below, in which the Mach number of
the flow at each characteristic intersections is plotted as a function of the area
at that cross section (normalized by the inlet area where the flow is sonic).
This forms a dense cloud of points, since there are a large number of
characteristics being used. The solid line is the relation for one-
dimensional flow from Chapter 5. Note the excellent agreement between the
two solutions. The agreement can be made even better if the channel area is
modified to vary more slowly. This result is a good validation of the method of
characteristics solver (or a good validation of one-dimensional flow, depending
on your perspective).
A method of characteristics program such as this can be used to solve more
complex problems where one dimensional solutions are not available, such as
the supersonic flow over a curved wedge. A solution to the hypersonic flow
(M =17.7) over a curved wedge whose surface is described by a power law
(y =0.2 x
0.8
) using the same program is shown below. Note that, due to the
*
Calculations in this section performed by Patricia Vu, McGill University.
flow
A
A
*
------
Compressible Fluids Notes
160
curved shock wave, the flow is no longer irrotational. Thus, the solver needed
to track an additional characteristic (the particle paths) in order to account for
the nonuniform entropy of the flow. This required a minor modification to the
solution techniques developed in this chapter.
s
h
o
c
k
w
e
d
g
e
s
u
r
fa
c
e
flow